igm emission observations with the cosmic web …authors.library.caltech.edu/44109/1/1402.4816.pdf2....

of 62

IGM Emission Observations with the Cosmic Web Imager: I. The Circum-

QSO Medium of QSO 1549+19, and Evidence for a Filamentary Gas Inflow

D. Christopher Martin1, Daphne Chang1,2, Matt Matuszewski1, Patrick Morrissey1, Shahin

Rahman1, Anna Moore3, Charles C. Steidel4

1Cahill Center for Astrophysics, California Institute of Technology, 1216 East California

Boulevard, Mail Code 278-17, Pasadena, California 91125, USA. 2Deceased. 3Caltech Optical

Observatories, Cahill Center for Astrophysics, California Institute of Technology, 1216 East

California Boulevard, Mail Code 11-17, Pasadena, California 91125, USA. 4Cahill Center for

Astrophysics, California Institute of Technology, 1216 East California Boulevard, Mail Code

249-17, Pasadena, California 91125, USA

Running Title: Cosmic Web Imager Observations of Gas Near QSO 1549+19

Accepted to the Astrophysical Journal

of 62

Abstract

The Palomar Cosmic Web Imager (PCWI), an integral field spectrograph designed to

detect and map low surface brightness emission, has obtained imaging spectroscopic maps of

Lyα from the circum-QSO medium (CQM) of QSO HS1549+19 at redshift z=2.843. Extensive

extended emission is detected from the CQM, consistent with fluorescent and pumped Lyα

produced by the ionizing and Lyα continuum of the QSO. Many features present in PCWI

spectral images match those detected in narrow-band images. Filamentary structures with narrow

line profiles are detected in several cases as long as 250-400 kpc. One of these is centered at a

velocity redshifted with respect to the systemic velocity, and displays a spatially collimated and

kinematically cold line profile increasing in velocity width approaching the QSO. This suggests

that the filament gas is infalling onto the QSO, perhaps in a cold accretion flow. Because of the

strong ionizing flux, the neutral column density is low, typically N(HI ) 1012 −1015cm−2 , and

the line center optical depth is also low (typically τ 0 <10 ), insufficient to display well-separated

double peak emission characteristic of higher line optical depths. With a simple ionization and

cloud model we can very roughly estimate the total gas mass ( logMgas = 12.5 ± 0.5 ) and the total

( logMtot = 13.3± 0.5 ). We can also calculate a kinematic mass from the total line profile (

2 ×1013M ), which agrees with the mass estimated from the gas emission. The intensity-binned

spectrum of the CQM shows a progression in kinematic properties consistent with heirarchical

structure formation.

of 62

1. Introduction

Theories of cosmological structure formation in a Cold Dark Matter (CDM) dominated

universe predict that the matter in the intergalactic medium (IGM) and the galaxies formed from

it are organized in a "cosmic web" of walls and filaments (Bond, Kofman, & Pogosyan 1994;

Davis et al. 1985; Frenk et al. 1985; Miralda-Escude et al. 1996) – with structures spanning

scales ranging from about one megaparsec to a few gigaparsecs (Mpc, Gpc, comoving). The

intersections of sheets and filaments mark regions of high overdensity in which galaxies form

and thrive, while the vast spaces between them are devoid of matter in comparison. Galaxy

redshift surveys uncovered this morphology on large scales several decades ago (de Lapparent,

Geller, & Huchra 1986). The gas associated with the filaments has been probed for nearly half a

century via absorption features in the spectra of high redshift quasars (Gunn & Peterson 1965;

Lynds 1971). A comprehensive review of the derived IGM properties is given in Meiksin (2009).

These investigations have revealed the evolution of the IGM, its enrichment with metals and

association with galaxies. For instance, the spectral density of the Lyman alpha (Lyα) forest

lines at redshift z ~ 3 indicates that nearly all baryons in the universe at that time were in a

relatively cool phase (T ~ 104K) (Rauch et al. 1997). At lower redshifts, near the current epoch,

the Lyα forest thins out considerably with a large fraction of the cool gas being shock heated as

it undergoes further gravitational collapse within the filaments, giving rise to a warm-hot

intergalactic medium (WHIM) (Bregman 2007). Some of this gas accretes onto the virialized

dark matter halos coalescing onto galaxies. Numerical simulations of structure formation are

consistent with this picture, suggesting that cosmic web forming baryonic matter traces the

underlying dark matter well (Cen et al. 1994; Cen & Ostriker 2006; Fukugita & Peebles 2004;

Hernquist et al. 1996; Prochaska & Tumlinson 2009; Zhang, Anninos, & Norman 1995). The

formation and evolution of galaxies is expected to be intimately connected to the IGM. Recent

modeling indicates that galaxies may acquire a large fraction of their stellar fuel via streams of

cold gas flowing along IGM filaments. This gas can fuel rapid periods of star formation and

could explain the abundance of starburst galaxies observed around z ~ 2 (Birnboim & Dekel

2003; Faucher-Giguère, Kereš, & Ma 2011; Kereš et al. 2005; van de Voort et al. 2011a; van de

Voort et al. 2011b). The computer models are being continually refined, however, resulting in

of 62

modifications of the predictions of the relative impact of “cold” and “hot” mode gas accretion

(Nelson et al. 2013). Observations of the morphology and kinematics of the IGM around galaxies

and its correlation and co-evolution with them are necessary to vet the computer models and to

further test our understanding of cosmic structure and galaxy formation and evolution (Faucher-

Giguère & Kereš 2011; Fumagalli et al. 2011; Kirkman & Tytler 2008). While QSO absorption

sightlines have yielded volumes of information about the co-evolution of IGM gas and galaxies,

the information is limited in nature due to the sparseness of suitable background continuum

sources on the sky and is, largely, statistical. Probing the full three-dimensional structure of

specific systems requires another approach – either relying on a different population of

background sources (more numerous or extended; nominally galaxies) or measuring emission

from the systems of interest. The former is likely not feasible until the 30 meter class telescopes

come online, while the latter is within the capabilities of current instrumentation, particularly in

the case of systems developing in regions of high overdensity.

Emission in strong UV resonance lines, particularly HI Lyα1216Å, is predicted to

delineate the cosmic web (Binette et al. 1993; Cantalupo et al. 2005; Furlanetto et al. 2003, 2005;

Gould & Weinberg 1996; Hogan & Weymann 1987; Kollmeier et al. 2010). For the vast

majority of gas within the cosmic web the emission is expected to be faint (~10-20

erg/s/cm2/arcsec2), as the only source of ionizing radiation is the relatively weak metagalactic

UV background. Lyα fluorescence from optically thick IGM illuminated by strong ionizing

sources, typically QSOs, is thought to be boosted above the typical level by several orders of

magnitude (Cantalupo, et al. 2005). Another energy source anticipated to fuel detectable

emission that needs to be taken into account when characterizing circum-galactic environments

is collisional excitation due to gravitational infall of gas, for example from cold accretion flows

(Fardal et al. 2001; Faucher-Giguère et al. 2010; Förster Schreiber et al. 2006; Haiman & Rees

2001). The expected surface brightness from the infalling gas is estimated to be on par with

Lyα fluorescence due to QSO ionizing radiation (Dijkstra, Stuart, & Wyithe 2006).

Significant effort is being exerted to detect and characterize the IGM in emission. Deep

narrowband observations (Fynbo, Moller, & Warren 1999; Hayashino et al. 2004; Nilsson et al.

2006; Steidel et al. 2000) have imaged numerous Lyα blobs in overdense regions while

of 62

Cantalupo, Lilly, and Haehnelt (2012) report a detection of a rich field of Lyα sources around a

z~2.4 quasar. Various techniques, both imaging and spectroscopic, have been used to detect

Lyα halos around galaxies and quasars (Francis & Bland-Hawthorn 2004; Hennawi et al. 2009;

Hu & Cowie 1987; North et al. 2012; Steidel et al. 2011). Long- and multi-slit spectroscopic

investigations have resulted in detections of a number of Lyα emitters in the vicinity of quasars

(Rauch et al. 2008). More recently, spectroscopic surveys have yielded intriguing observations of

Lyα emission from the IGM around quasars and galaxies exhibiting possible filamentary nature

and velocity structure (Hennawi & Prochaska 2013; Rauch et al. 2013). There have also been

reports of emission associated with damped Lyα systems near proximate QSOs, although the

observed flux is thought to originate from either the DLA host galaxy or the background quasar

(Adelberger et al. 2006; Hennawi, et al. 2009). The evidence appears strong that volumes around

galaxies and quasars are rich in gas that manifests a complex morphology, kinematic structure,

and interacts extensively with the central object. Unfortunately, the above-outlined observations

and techniques do not result in a full picture of the individual systems being studied, yielding

either incomplete spatial or limited spectral information. Integral field spectroscopy, however,

combines moderate resolution spectral content with spatial data over a compact field that is well

matched in shape and size to the circum-galactic and circum-QSO environments. As such, it is a

great tool for characterizing galaxies and their neighborhoods. We have initiated a program to

study faint emission around galaxies, quasars, and Lyα nebulae with the Palomar Cosmic Web

Imager, an integral field spectrograph for the Hale telescope at Palomar Observatory built

specifically to detect and map dim and diffuse emission.

In this paper we report the on the detection and characterization of extended Circum-

QSO Medium (CQM) gas surrounding QSO HS1549+1919 with PCWI. In §2 we describe the

instrument, data and analysis. In §3 we compare the PCWI results to the narrow-band image of

the QSO. In §4 we describe our smoothing algorithm (3D Adaptive Smoothing including

Lambda [3DASL]), which we use to display image slices of the observed spectral data cube.

Section §4.2 is a detailed technical discussion of robustness tests for this algorithm and the

results. We show a number of simulations and additional information which we hope builds

confidence in the applicability and fidelity of this algorithm to moderate signal-to-noise spectral

image cubes. This section can be skipped without loss of continuity. In §5 we study the

of 62

morphology and derive physical parameters from the data and simple models. We summarize the

conclusions in §6. We use a standard WMAP-7 cosmology (Komatsu et al. 2011), and unless

otherwise specified all distances are given in proper (physical) coordinates.

2. Cosmic Web Imager Data and Analysis

2.1. The Palomar Cosmic Web Imager

We have constructed an integral field spectrograph called the Palomar Cosmic Web

Imager (PCWI), that is designed to search for, map, and characterize IGM emission and other

low surface brightness phenomena (Matuszewski et al. 2010). It uses a 40" x 60" reflective

image slicer with 24 40" x 2.5" slices. For these observations the spectrograph and slicer was

oriented with individual slices running east-west. The spectrograph has a high dispersion volume

phase holographic grating covering 4500-5400Å with an instantaneous bandwidth of 400Å. The

spectrograph attains a slit-limited spectral resolution of Δλ~1Å and a peak efficiency of ~10% at

5000Å including the telescope and atmosphere. PCWI is mounted at the Cassegrain focus of the

Hale 5 meter telescope on Mt. Palomar. As a spectroscopic imager it is possible to form

individual 1Å wide images of the sky. The imaging resolution, while limited by the 2.5" slicer

sampling, can be effectively improved to ~1.3" by dithering the field between individual

exposures.

2.2. Observations of QSO HS1549+19

Since it is likely that IGM emission is brightest in regions near bright QSOs, we observed

the region around QSO HS1549+1919 (z=2.843) that is known to harbor extensive narrow-band

Lyα emission (Trainor & Steidel 2012). Figure 1 shows a narrow-band Lyα image obtained at

Keck.

We obtained a total of 6.5 hours on-source and 6.5 hours off-source exposure centered on

QSO HS1549+19 (15h51m52.48s, +19°11´4.2˝), over a total region 120 arcsec East-West by 60

arcsec North-South. Individual exposures were obtained using a nod-and-shuffle technique

developed for long-slit observations (Sembach & Tonry 1996). We used the central 1/3 of the

CCD for recording the spectrum, and masked the outer 2/3 of the CCD (restricting the spectral

bandpass to ~150Å). Each frame was created using the following proceedure. The telescope is

of 62

pointed at the source region, the guide camera locks onto a guide star, the shutter is opened, and

a two-minute exposure is taken. Then the shutter is closed, the telescope nodded to a background

location 1.5 arcminutes north and 2 arcminutes west, the CCD is clocked by 670 pixels up, the

shutter is opened, and a background image is obtained. In order to make an unbiased background

subtraction the precise sequence is 1 minute background, 2 minutes source, 2 minutes

background, …, 2 minutes source, 1 minute background for a total of 11 background and 10

source exposures, each a total of 20 minutes. A total of 10 of these source/background cycles

was performed (a total of a little more than 40 minutes with overhead), followed by CCD frame

readout. A total of 14 20/20 minute source/background and 11 10/10 minute exposures were

obtained over 6 observing nights in 2011 and 2012. The objective of this method is to provide

the highest precision sky subtraction possible without increasing the read noise. The CCD was

read-out in a 2 by 2 binning mode to reduce the impact read noise further. The CCD read noise

was less than 3 e- (rms).

The PCWI field of view is 40 arcsec (parallel to slices) by 60 arcsec (sum of 24 2.5

arcsec wide slices). This observation was generated from a mosaic of multiple pointings covering

a total region of 120 arcsec by 60 arcsec. During this observation PCWI was oriented with slices

in the east-west direction. In some regions of the mosaic several north-south pointings were

dithered by fractional slit widths to provide the advantage of further averaging over instrument

features and slice-to-slice efficiency variations. Background pointings were dithered typically by

0.5-2 arcsec. The background area was chosen because it has minimal bright narrow-band or

continuum sources. We also obtained numerous calibration images using lamps throughout the

observing period.

2.3. PCWI Pipeline Analysis

The PCWI pipeline converts CCD images into a 3D data cube. All individual exposures

are bias-subtracted using a superbias, and cosmic rays are removed. Detector images are then

split into source and background regions. Continuum flats are used to define the slice regions on

the detector. An edge calibration image using a continuum lamp corrects is used to refine the

horizontal registration of each CCD row, and a pinhole image is used to fix the plate scale in

each slice. An arc lamp spectrum is sliced and the line tilts measured to correct for spectral line

of 62

curvature in each slice. The resulting dewarped spectrum is then calibrated against the standard

input spectrum to yield a polynomial slice-by-slice wavelength solution.

Science images are sliced, edge rectified, rescaled, dewarped, and wavelength rectified

into source and background slice spectra. They are wavelength-shifted to compensate for <1Å of

total flexure using sky lines. The final mosaiced data cubes are generated using the astrometry

based on the guide star camera, and astrometry was checked using several moderately bright

objects in the source fields. Astrometry is good to about 0.5 arcsec (rms). Mosaiced data cubes

were created for source and background. Source data cubes are rectified based on astrometry,

and background data cubes were created using the same source data cube registration (so that

individual source and background exposures are subtracted one-to-one). Exposure maps are

generated by first generating a smoothed “skyflat” from background spectral images in detector

coordinates. These are normalized, exposure-time multipled, and mosaiced in the same way as

source and background images. The result is a set of flux data cubes (RA, DEC, and λ) for each

exposure, sampled at (0.292", 0.292", either 0.25Å or rebinned to 1Å) covering 4600-4750Å.

The difference cube is obtained by simply subtracting source and background flux cube. Because

the nod-and-shuffle mask does not physically contact the CCD, a small amount of diffuse

continuum light remains in the subtracted cube (<1%), which is easily removed with a low-order

continuum fit.

In the resulting difference cube, structures spanning 10" x 10" x 4Å have a 1-sigma noise

level of ~1400 ph cm-2 s-1 sr-1 Å-1 (Line Units/Å = LU/A). Sky levels were ~1 x 105 LU/Å (ph

cm-2 s-1 sr-1 Å-1). Our emission line detections are at the level of ~3-10% of sky. Sky levels

typically varied smoothly and mostly linearly by 0.5-2% on 20 minute timescales or typicaly

0.1% on the 2 minute nod-and-shuffle cadence, leading to sky-subtraction errors significantly

less than 300 LU/Å for individual exposures, and correspondingly smaller errors for the summed

exposures. 1000LU corresponds to ~10-19 erg cm-2 s-1 arcsec-2 at 5000 Å.

3. Comparison of PCWI data to the Narrow-Band Image

We would like to understand to what extent features detected in the PCWI data are

present in and confirmed by the narrow-band image. In Figure 1 we show the narrow-band image

obtained at Keck, along with broad and narrow slice images obtained with PCWI. The narrow-

band image is centered on 4670Å with an 80Å width. Extensive emission is seen near the QSO,

of 62

generally of a clumpy appearance. We have outlined several of these clumps in Figure 1a and in

the other figure panels. In Figure 1b we show a broad (32Å) slice sum obtained from PCWI. The

sum is produced using the smoothed data cube described in the next section. Clumps

corresponding to the three most prominent clumps in the narrow-band image are observed in

PCWI as well, albeit at lower spatial resolution and therefore with less distinction. In Figure 1c-d

we show narrower PCWI slices centered on two prominent emission features which we later

label as filament 1 (easterly extension centered at 4682Å) and filament complex 2 (westerly

network of extended emission centered at 4672Å). In narrow slices the emission becomes more

extended and filamentary and of lower intensity. In both cases the clumps are at the base (with

respect to the QSO) of the more extended, narrow spectral width extensions. As we show later

the clumps are much broader than the extended filaments. The narrow-band image is better at

detecting spectrally broad emission clumps while PCWI is better at detecting extended,

spectrally narrow emission features. Spectral confusion and to a lesser extent the increased noise

in the broader band narrow-band image makes it less ideal for detecting narrow, filamentary

features, while the clumps appear less distinct in the PCWI data because of lower resolution and

greater noise in the wide bin slice. Because of the detection of extensive, narrow emission in the

eastern third of the mosiac near the systemic wavelength (filament complex 2), we have added

additional regions in order to make a quantitative comparison with the narrow-band image

(shown in Figure 1a and 1b) in order to test for a spurious origin for filament complex 2.

In Figure 2 we show a comparison of the emission in these regions between the PCWI

and narrow-band fluxes. The narrow-band image has a nominal Poisson noise level of 2.5 x 10-19

erg cm-2 s-1 arcsec-2 (2500LU) with a 1 arcsec2 smoothing box. Because of the narrow-band filter

width (80Å), typical sky (50,000LU/Å) corresponds to a total level of 4 x 106 LU. A 0.1% error

in continuum subtraction would lead to a 4000LU error in the corrected narrow-band image.

Continuum-subtraction errors in the narrow-band images are introduced by a number of factors,

including the spatial/spectral variation in the instrument response over the narrow vs. broad

continuum filter, the variation in the sky spectrum over the two wavelength ranges, the impact of

continuum source wings and scattered light, and the additional narrow-band objects present in

the broader continuum image. Spatially extended structures are particularly subject to these

errors. We estimate from variations in the zero level (which occasionally goes negative) that

typical systematic continuum-subtraction errors are at least ~0.1-0.2% of sky, or ~4-8 x 10-19 erg

of 62

cm-2 s-1 arcsec-2 (~4000-8000LU), depend on the spatial scale, and may be higher in some

regions. There are large, low spatial frequency variations that lead to the flux offset of ~10 x 10-

19 erg cm-2 s-1 arcsec-2 (~10,000LU) apparent in Figure 2. The CWI image summed in ~5 x 5

arcsec2 boxes (Figure 1b) has a similar Poisson error level (~10000LU). Thus for both images

typical noise amplitude will cause variations from blue to green color values on 5 arcsec scales,

as is observed. Comparison of individual features detected in either image will not be one-to-one

because of this noise level. We note that the PCWI image has not had continuum sources

removed, while the narrow-band image is continuum-subtracted. Given the uncertainties, the

fluxes measured in the various regions are consistent within statistical and continuum subtraction

errors.

To confirm the spectral content of these regions and the fidelity of the smoothing

algorthm in reproducing this spectral/spatial content, we show the spectra obtained in the blob

regions outlined in Figure 1 in Figure 3 using the unsmoothed data cubes. Region a shows a

narrow component at 4683Å, consistent with the redshift derived from Keck/LRIS follow-up

observations of the clump. Region b shows a broad component centered on 4673Å, matching the

redshift derived from Keck follow-up on this clump. Region c shows a moderate width feature

centered on 4670Å, again consistent with the redshift derived from Keck/LRIS follow-up.

Finally, region e, which does not appear as a distinct emission component in the narrow-band

image, but does appear as a filamentary connection between clump b and c in the PCWI 4672Å

slice, shows a narrow component at 4671Å. This corresponds to the systemic velocity of the

QSO. All of the observations are consistent with a trend analyzed in § 5.3 in which faint

emission is spectrally narrow and more extended, while spectrally broad emission is associated

with the brighter emission clumps. The total region around the QSO is displayed in the final

panel, and a summary of fluxes, luminosities, and detection significance of the detected emission

features is provided in Table 1. The emission is detected with high significance in all the regions.

We conclude that, taking into account the nature of the underlying emission and the

relative strengths of the two observational techniques in detecting spatially extended, spectrally

narrow emission, and spatially compact, spectrally broad emission, that the two observations are

consistent.

of 62

4. Adaptive Smoothing Algorithm

In general, even long exposure data from PCWI targeting IGM emission is of modest

signal-to-noise ratio. IGM emission models predict that line emission will be extended, probably

filamentary, and may show significant kinematic variations with spatial position (Cantalupo, et

al. 2005; Kollmeier, et al. 2010). We have therefore developed and extensively tested an adaptive

smoothing algorithm designed to highlight extended line emission while not artificially

smoothing more compact emission line and continuum sources.

4.1 Results of Adaptive Smoothing Algorithm

The slice images displayed in Figure 4 (and following) are produced from the raw

difference flux cube by a 3D adaptive smoothing algorithm (3DASL). The algorithm works as

follows. The (0.292" x 0.292" x 1Å spaxel) data cubes are first boxcar smoothed spectrally by

2Å, then smoothed spatially (boxcar or Gaussian kernels), by successively larger smoothing

kernels ranging from FWHM=3 pixels (1 arcsec) to 70 pixels (20 arcsec) in 2 pixel increments.

Boxcar or Gaussian kernels produce similar results, but the latter exhibit less high-spatial-

frequency noise. The effective signal-to-noise threshold is correspondingly reduced by a factor

equal to the one pixel noise divided by the 1D smoothing width (e.g., a 1-sigma 1-pixel noise of

9000LU becomes 1800LU for a 5 pixel smoothing box, and 180LU for a 50 pixel box).

Whenever a spaxel exceeds the noise threshold (which of course depends on the kernel) it is

placed in the final cube (detected). The flux in the smoothed spaxel is subtracted from the

unsmoothed cube pixel prior to the next spatial smoothing cycle. Any residual flux in the spaxel

can still contribute to the smoothed slice images but is not available for further “detection”. The

unsmoothed difference cube is then smoothed in wavelength by a 4Å kernel, and the spatial

smoothing algorithm is repeated. This loop repeats once more with a spectral smoothing of 8Å,

or a total of three times. The 2.5-sigma noise threshold was derived directly from the difference

cube, and is consistent with the predicted Poisson noise. We discuss below variations in the

algorithm and signal-to-noise thresholds.

We show in Figure 4 spectral images in 4Å slices obtained with this algorithm. We show

in Figure 5 images that illustrate the raw data and noise in comparison to the adaptively

smoothed image slices. We show later that pure noise produces a very different smoothed image

of 62

and emission morphology. Many of the features apparent in the spectral images appear in

multiple 1Å slices, as we show in Figure 6 and Figure 7.

4.2 Robustness Tests of the Adaptive Smoothing Algorithm

It is important to note that the principle role of the smoothed images is to identify

extended emission in order to define regions that can be summed and studied, typically with the

unsmoothed difference cubes. Because of the modest signal-to-noise ratio of the emission there

will not be a detailed one-to-one correspondance of the smoothed slice images and the

underlying data, particularly on small scales. As we show later with simulations, the algorithm

tends to produce less extended and more clumpy emission than is underlying, again due to

modest signal-to-noise. However as we demonstrate below there is a good correspondance on the

scales of interest for this investigation.

1. Simple filament simulations. We have performed simulations to investigate the

behavior of the algorithm under various conditions. We first show a set of simulations that are

performed for a 300 x 300 pixel image (single wavelength slice). Figure 8 shows the first set of

simulations which consist of a set of compact sources distributed in a “filamentary” pattern.

2DASL does a reasonable job isolating the sources, while simple smoothing tends to produce an

artificial extended filament. In Figure 9 we add a long, narrow emission filament. The algorithm

clearly revels extended, contiguous emission, although noise introduces small-scale structure in

the filament which is not real. In Figure 10 we complicate the filament by adding a narrow core

and broad wings. Again, 2DASL does a reasonable job revealing this more complex extended

structure, while isolating the compact sources. In all three cases a mean of 10 realizations is quite

close to the original image, showing that the while each individual image is still noisy, the mean

of many images converges to the input image.

2. Modifications to the signal-to-noise threshold and addition of bright central

source. For a pixel to be detected it must exceed the signal-to-noise threshold. The choice of the

signal-to-noise threshold has a signficant effect on the resulting smoothed images. Our extended

emission detections are modest in signal-to-noise ratio, so we have investigated the impact of

this threshold on the resulting smoothed images. We show in Figure 10 and Figure 11 the impact

of the threshold on the 2D image simulations. The presence of a very bright central source could

impact the algortim, so we show in Figure 12 a simulation with such an added source. In this

of 62

case the central source is not removed prior to smoothing, in order to check for any impacts.

There is none outside a central zone a few arcseconds in radiuss.

In Figure 13 we have simulated a series of filaments with increasing intensity. This

illustrates a detection threshold corresponding to (S/N)*σ/θ, where (S/N) is the smoothing

algorithm threshold, σ is the noise level (per pixel), and θ is the minimum feature size (narrow

width of filament). In Figure 14 we show how the 3DASL results for real data in 4Å slices varies

with S/N threshold from S/N>3.0 to S/N>2.0. Lowering the threshold has the principle effect of

increasing the number of compact noise spikes detected by the algorithm, and very slowly

increasing the size of extended features (real and noise). Based on this study we conclude that a

conservative compromise is a threshold of S/N>2.5.

3. Filament simulation including realistic noise, line profiles and velocity variations.

PCWI spectral images do not produce square image pixels, and are the result of many dithered

individual exposures. This is because the coarse slice width (2.5 arcsec) undersamples the seeing

PSF, and a single detector pixel is remapped to a 2.5 arcsec by 0.6 by 0.25Å arcsec spaxel in the

data cube. We performed simulations of a data cube using the measured sky spectrum and

spectral/spatial distortion to produce an input image (including source and background detector

regions) with Poisson noise. Data cubes were generated corresponding to each QSO HS1549+19

exposure time and position using the standard pipeline. The individual exposure cubes were

coadded in a mosaic and the source and background cubes differenced, following the identical

steps as the actual data. The resulting raw, smoothed, and 3DASL smoothed simulated image

slices are displayed in Figure 15 and compared to actual data. This figure shows that most of the

noise is manifested as spurious compact emission regions, and only very occasionally is a low-

level extended feature generated by random, contiguous regions that have spatial correlation

partially produced by the tall, narrow spaxels. Comparing these images to Figure 4 shows that

the detected extended features are not a result of spurious smoothing of noise introduced by the

3DASL algorithm.

We add a filament with a flux level comparable to that detected in QSO HS1549+19

environment. We first assume the emission line has a width of 3Å and is fixed in velocity. The

image slice at the line wavelength is given in Figure 16. The filament is clearly detected in the

4Å image slice. We show a “spectral-image” plot, in which we have sliced the smoothed data

of 62

cube parallel to the filament axis and the wavelength direction, in Figure 17. The spectral image

is summed over a “pseudo slit” of width 5 arcsec. This is a method to show how individual

structures that may show kinematic variations along their length, which would render them as

disconnected structurs in distinct image slices, may form a single extended feature. Simulations

suggest that velocity variations could be hundreds of km/s along single filaments (Cantalupo, et

al. 2005), while a single 1Å slice is 64 km/s wide. Finally, we added a velocity shear to the

filament similar to that predicted in cosmological simulations. The resulting spectral-image plot

also in Figure 17 shows that this shear can be detected and does not affect the detectability of the

filament.

We also studied the effects of the dither pattern in each area of the mosaic. It can be seen

from Figure 5 that the western third of the image mosaic has lower vertical resolution,

comparable to the slice width 2.5 arcsec. Because of an inadvertant axes swap, these images

were obtained without a half-slice north-south dither (perpendicular to the slices) used in the

other 2/3 of the cube. The impact of this on the effective point spread function is to change the

north-south 1D PSF from approximately a triangle function with FWHM = 2.5 arcsec to a square

function with full width 2.5 arcsec. There should be no significant impact on the detectability of

individual features broader than the slice width in the north-south direction, and very little

additional susceptability to systemic error producing false detections than in the areas with a 1.25

arcsec north-south dither, since source and background data cubes are obtained in the identical

fashion, and the dither is limited to 1.5 slices and therefore does not significantly reduce artifacts

produced by an improperly flat-fielded single slice. We demonstrated this by generating a

simulated filament extending to the west with intensity comparable to that detected in the actual

data (corresponding to filament 2). The slice image and the spectral image are displayed in

Figure 18.

4. Features are detected in multiple slices. We note that features detected in the data

are clearly detected in multiple image slices. We identify two filaments in Figure 4, as discussed

below in §5.1. These filaments are denoted filament 1 and filament complex 2. Filament

complex 2 (centered on 4673Å) is detected in at least four 1Å slices (Figure 6). Filament 1

(centered on 4682Å) is detected in at least seven 1Å slices (Figure 7). While we expect some

slice to slice correlation due to the wavelength component of the smoothing algorithm, pure

noise does not produce such correspondance over many slices.

of 62

5. Compact sources--impact on results. A large number of faint compact sources could

in principle be converted by the smoothing algorithm into apparent extended emission in the

smoothed images. Such sources if unsubtracted could also add spurious emission to the

unsmoothed data cube and spectra derived from regions. We have therefore carefully examined

the impact of compact sources on the results.

We showed in Figure 8-10 that compact sources resolved by the instrument are also

resolved by the smoothing algorithm, and in contrast to simple smoothing do not, as a rule, form

spurious extended features. Of course a high surface density of compact sources would not be

resolved by the instrument, and we cannot rule out that extended features are formed by the

summation of many individual clumps or sources of Lyα emission. The narrowband image

(Figure 1) suggests that there are emission-line clumps but that underlying extended components

could also be present. Both of filaments 1 and 2 seem to be associated with clumps in the

narrowband image that also are present in the PCWI slices. The clump associated with filament 1

appears in panel h of Figure 4, at 4984Å. The clump associated with filament complex 2 appears

in panels c and d (4664 and 4968Å).

Continuum and line sources can be removed from the data cube to test their impact. We

used a continuum image to determine the location and approximate brightness of all compact

continuum sources in the field of view of our source field. We determined the point spread

function (PSF) from prior observations. In this PSF 80% of the light is confined to a region of

7.4 arcsec2. We used a source removal algorithm that uses this PSF and neighboring sky, slice by

slice, to remove each source and replace it with an estimate of local sky (which could include

local extended emission). There are a total of 84 sources detected in the continuum image of the

source field to V<25. Of these, 18 are Lyα sources in the narrow-band image, which we did not

remove. The remaining 66 continuum sources are removed. Figure 19 shows the compact source

distribution and magnitudes superimposed on the narrow-band image, along with the boxes used

for source definition. There are no significant differences in the resulting smoothed images or

region spectra with or without source subtraction. This figure also shows an image comparison

with and without sources subtracted. All images and spectra are given without sources

subtracted.

of 62

5. Properties of Extended Emission

5.1. Morphological Properties

We begin by describing the distribution of emission as displayed in the image slice plots

of Figure 4, Figure 6, and Figure 7. The emission is most prominent in the band 4660-4685Å. In

slice 4664Å a large exension is present toward the north-west. Moving from 4664A to 4672Å the

extension moves southward and grows to fill the western portion of the field. The cloud at Δα=-

30, Δδ=-7, prominent in Figure 1, is clearly present in slice 4672Å. Clumps nearer to the QSO on

the west side in Figure 1 are also visible here. The cloud and the QSO are bridged by emission

along the east-west direction, as well as emission looping southward. In slice 4676Å the

emission extends mostly to the south. In slice 4680Å a narrow filament extends to the east and

slightly south, as far as 30 arscec from the QSO. In 4684Å the filament is more of a clump, a

feature also present in Figure 1. Finally in 4688Å the extended emission has mostly vanished.

We described in the previous section how spectral-image plots can delineate what could

be contiguous extended features exhibiting kinematic shear. We have generated an array of these

plots in pseudo-slits centered on the QSO with varying position angle, shown in Figure 20. The

extensive extended emission near the systemic velocity and apparent in Figure 4 is also present

in these plots. While extended emission is seen at all azimuths, regions show strong, narrow

emission lines. Azimuth 345° shows a strong (#1, cf. below) near 4682Å. Azimuth 180°-210°

(shown as Δθ<0 for azimuth 0°-30°) shows extension out to r>50 arcsec near the systemic

velocity.

In Figure 21 and Figure 22 we use the image slice to trace out pseudo-slits guided by the

image morphology. Figure 21 shows filament 1 extends to at least r=30 arcsec, and may extend

to greater than 40-50 arcsec. Double peaked emission could be present between 20 and 40

arcsec. The feature shows a central velocity near 4682Å, with modest velocity shear along its

length, until r=12 arcsec. At this point the velocity shifts sharply to approach the QSO systemic

value at 4672Å. This point also corresponds to a large clump in the narrowband image.

The emission complex appearing mainly around slice 4672Å can be described as

extensive, compared to that in most of the other slices, and perhaps “web-like”, with clumps

of 62

embedded in a network of apparent filaments. We have delineated one path through these

filament-like connections in Figure 22, which we denote filament 2. The extension in Figure 22

bridges the QSO with the large region of clumpy emission in the west. The emission is narrow

and near the systemic velocity for r>27 arcsec. In the zone 20<r<30 arcsec the emission profile

broadens sharply and is accompanied by continuum, at the location of a prominent line and

continuum source. While we showed that that narrow-band image and CWI data are roughly

consistent, because this complex does not have a clear counterpart in the narrow-band image, we

regard this detection as more tentative than filament 1.

5.2. Physical Properties

We use a simple CLOUDY model (Ferland 1996) to derive physical properties of the

CQM. We use the known QSO HS 1549+19 brightness, normalized at Lyα, of

LQ = 1×1048 erg/s , and a standard assumed spectral energy distribution (AGN T=1.5 x 105K,

a(ox)=-1.4, a(uv)=-0.5, a(x)=-1) to derive the ionizing and Lyα photon-pumping continuum. The

Lyα line is extremely broad and the continuum normalization includes it. For the reference

model, we assume isolated clouds with a constant physical thickness (along the radial direction

to the QSO) equal to 40 kpc at all radial distances (corresponding to 5 arcsec). The cloud

distance from the QSO is variable. The clouds are assumed to have a constant density that is a

free parameter, unit filling factor, and metallicity 10% solar to approximate abundances in

circum-QSO gas. Dust absorption is ignored. The total intensity in Lyα is calculated for each

radial cloud location, in a +/-20Å window around the systemic Lyα wavelength of 4672Å. The

radial location and total intensity then implies a cloud density, column densities of N(HI) and

N(H), mean gas temperature, etc.

We show the radial dependence of cloud density and intensity, compared with observed

values, in Figure 24. Three processes produce Lya photons: radiative recombination, photon

pumping, and collisional excitation. Because of the hyperluminous QSO, the gas is highly

ionized and hot even when relatively dense, neutral hydrogen column densities are low near the

QSO, and most Lyman continuum photons escape. The temperature run with radius is shown in

Figure 26. The temperature is far higher than that in the general IGM and Lyman alpha forest.

We investigate below whether the derived temperature impacts the conclusions. Based on the

of 62

CLOUDY model output, collisional excitation accounts for at most 1-10% of the total line

intensity, in spite of the high temperature, because the neutral fraction is very low. In most

regions the Lyα line emission is dominated by photon pumping, or scattering of QSO continuum

light in the Lyα line of the surrounding gas. In Figure 24 solid lines give intensity vs. radius for

pumping and recombination (also called Lyα fluorescence or LAF). Dotted lines give the

pumping contribution. Lines that are pumped will not typically show a double peaked profile,

which arises when recombination photons are escaping a large optical depth in the line through

the wings. The vertical line shows the minimum inner radius of the analysis, corresponding to 10

arcsec (78 kpc). All masses are calculated using emission only outside this radius. Inside this

radius the wings from the QSO are too strong to allow secure conclusions. Note that we have

also not included any gas cooling due to dissipation of gravitational energy as it accretes into the

dark matter halo. This approximation is accurate for the total emission in the field of view, which

has a flux of 1.7 ×10−14 erg s-1 cm-2 (cf. Table 1) about 500 times higher than the flux predicted

for a 7 ×1012M halo (Haiman, Spaans, & Quataert 2000).

From this reference model analysis we can therefore derive an approximate gas masses in

the CQM. For filament 1 we derive a baryon (gas) mass of Mgas ≈1.2 ×1011M . For the

tentatively detected filament 2 we derive a baryon (gas) mass of Mgas ≈ 3.6 ×1011M . In each

case the mass is calculated by summing the data cube over a range λSYS − Δλ < λ < λSYS + Δλ ,

where Δλ is 5-10Å for filaments and 20Å for the total field. We then convert the total intensity

to a mass density using the model, and summing over either the filament or the full data cube not

including the central 10 arcsec radius region around the QSO (area 3700 arcsec2). The total Lyα

luminosity of the observed data is LLyα ≈1.3×1046 erg/s , or 1.3% of the total QSO luminosity.

Using our simple model we derive a rough estimate of the total baryon (gas) mass of

Mgas ≈ 2.9 ×1012M . This implies a minimum total mass (gas plus dark matter) of

Mtot ≈ 2.3×1013M . This mass estimate is not changed significantly when we use the source

subtracted data cube. At z=2.843, this total halo mass corresponds to a virial radius of rvir = 230

kpc (or 30 arcsec) and circular velocity of vc=650 km/s. As we see below, this is probably

consistent with the average gas kinematics. We derived this total mass using the S/N>2.0 data

of 62

cube. The corresponding values for thresholds S/N>2.5 and 3.0 are Mgas ≈ 2.5 ×1012M and

Mgas ≈ 2.2 ×1012M , indicating that less than 25% of the signal arises from 2.0<S/N<3.0

emission.

The total gas (baryon) mass of filament 1 and 2 is M1+2 ≈ 4.8 ×1011M , while the total

CQM gas (baryon) mass estimated above is M1+2 ≈ 3×1012M , a factor of 6 higher. The

definition of filament 2 is somewhat arbitrary since it is surrounded by an extensive emission

complex. This filament 2 complex represents a significant contribution to the total (~50%). Other

emission including a significant extension toward the north-north-west constitute the balance

(see Figure 4). The bulk of the emission is very significantly detected (76% has S/N>3), and as

we display in Figure 31 shows spectral characteristics consistent with velocity broadened Lyα

emission. These two estimates conservatively bracket the range of possible gas masses for the

CQM.

We now consider the uncertainties in the derived gas masses introduced by the model

assumptions. Clearly, the radiative transfer of the optically thick Lyα line leads to signficant

uncertainty and any kinematic trends which impact the line optical depth. However, because the

ionizing field from the QSO is so strong, the HI column density is moderate near the QSO. In

addition, the temperature of the gas is high. The combination of these two effects leads to

moderate line center optical depths (0.01≤ τ 0 ≤100 ), and in most regions τ 0 <10 , as we show in

Figure 25. Thus the number of scatterings is limited as is any absorption by dust in the CQM gas.

Other than dust absorption, radiative transfer effects will mostly redistribute the line photons into

the line wings and somewhat blur the correspondence between flux and local density (Kollmeier,

et al. 2010). But at these low optical depths the effects are likely to be modest. A careful

treatment of the effects of radiative transfer requires a more detailed geometrical and kinematic

model of the CQM, and will be addressed in a future paper.

The method of summing over wavelength represents an upper limit in the sense that

individual kinematic components are treated as a single slab. Since the intensity scales with

density squared, this estimate will be low by a factor equal to the square root of the number of

components. Typically for the total field the emission occupies a mean bandwidth of 4-8Å. If

of 62

there is one component per thermal broadening width (~2Å) then this underestimates the total by

a factor of 1.4-2 or an error of ~0.2 dex.

The next uncertainty is the location of the gas along the line of sight. However, it can be

seen from Figure 24 that the radial dependence is much weaker than the dependence on cloud

density. This is because the neutral fraction increases with the radius flattening the neutral

column density variation. In Figure 27 we show how conversion of intensity to intrinsic density

without knowledge of the radial distance leads to a typical error in ~0.3 dex. Also, the thickness

of the cloud (L) has a modest impact on the results: doubling the thickness produces a mass

estimate 0.1 dex higher.

The volume filling factor ( fV ) has a more significant effect on the mass estimate. All

three photon production processes are affected in a similar way by a less then unity fV . For

photon pumping, it is easiest to consider the case of an optically thin Lyα line. In this case,

IPP nLfV

nHn

⎛⎝⎜

⎞⎠⎟ g τ 0,b,T( ) .

where L is the thickness and n is the local density of the medium in the filled volume. The

function g accounts for optical depth and temperature effects. In photoionization equilibrium,

nHn

⎛⎝⎜

⎞⎠⎟ =

α RR

Γn ,

with so the resulting intensity scales as

IPP n2LfVg τ 0,b,T( ) .

For the conditions in the gas near the QSO the photoionization rate exceeds the collisional

ionization rate ( Γ nγ ) by >100 so this is a good approximation. This means that for a given

observed intensity, the density scales as n IPP1/2 fV

−1/2L−1/2 , and the total mass scales as

M nLfV IPP

1/2 fV−1/2L−1/2( )LfV IPP fV L( )1/2 . We find that when non-unity filling factor and

different cloud/filament thicknesses are incorporated into the model, including all processes, that

the inferred baryon mass is

of 62

Mgas 3.6 ×10

12 fV0.66 L

40kpc⎛⎝⎜

⎞⎠⎟

0.37

M .

There is no simple way to estimate the volume filling factor without resorting to detailed

physical simulations that probe small scales. We might expect fV <1 in a multiphase CQM,

because of the presence of some hot, virialized gas in the massive halo hosting the QSO which

can pressure confine cooler clouds responsible for the Lyα emission. Physical processes that

might lead to less than unity filling factor in the CQM include turbulent shear between inflowing

cold and hot accretion flows, multiphase QSO outflows, or the interaction of inflows, outflows,

and orbiting CQM gas. The clouds almost certainly would have to be pressure confined because

they are heated dramatically by the QSO and would expand in timescales much less than the

QSO lifetime if they are <1 kpc in size. Taking (somewhat arbitrarily) fV = 0.3 , the mass

estimate is lower by a factor of 2.

The metallicity assumption also impacts the predicted intensity moderately. Consider the

behavior at log R[cm]=23.5. If the metallicity is as low as 1%, the inferred density is increased

by less than 0.08 dex. If the metallicity is as high as solar, the predicted Lyα flux is decreased

and the inferred density increases. This effect is largest at higher density. At log n=-2.0 [-2.5],

the inferred density and mass would increase by 0.24 [0.06] dex, or 1.7 [1.14].

The temperature predicted by the CLOUDY model is high. Gas at this temperature would

produce very broad absorption (not seen in proximity systems) but because of the low optical

depth (typical equivalent widths less than 0.3Å) and width (b values of 50-100 km/s) the lines

could be difficult to detect. However, if the temperature is in error we could derive an incorrect

mass. To check this we fixed the temperature to T = 104.6K . We find that depending on the

filling factor the fixed temperature model is lower in mass by a factor of 1.2-1.4.

Finally we can estimate the total uncertainty in the mass estimate based on all these

factors. We take a fiducial filling factor of fV = 0.3 and we take the following estimated errors

for radial distance, component number, filling factor, thickness, and metallicity as σ R = 0.3 ,

σ n = 0.2 , σ f = 0.5 , σ L = 0.3 , σ T = 0.1 and σ Z = 0.1 (all in dex). Based on the model

sensitivities, this yields a total error of

of 62

σ M = 0.52 = 0.32 + 0.22 + 0.66*0.5( )2 + 0.36*0.3( )2 + 0.12 + 0.12( ) . The baryon mass estimate is

then logMgas = 12.5 ± 0.5 and the total mass is then logMtot = 13.4 ± 0.5 . This estimate is

tentative given the simple assumptions of the model. We return now to the two filaments

discussed above. Profiles of derived physical parameters are given in Figure 28 and Figure 29.

These plots give intensity, intrinsic density n, HI column density N(HI), total column density

N(H), temperature (from the model), velocity dispersion (from the line profile), and velocity

centroid. The figures show the virial radius. It is interesting to note that the velocity dispersion

for both filaments is low (~100 km/s) outside the virial radius, but increases inside to 200 km/s

(filament 1) and 300-400 km/s (filament 2).

Filament 1 is strongly detected out to 30 arcsec, and possibly detected as far as 45-50

arcsec in the S/N>2.0 adaptively smoothed image corresponding to 2-sigma (~1700LU) with a

feature linewidth ~1Å and structure size ~10x10 arcsec2 (see Figure 21). Therefor the filament

length is 230-400 kpc. The filament has a baryon mass of Mgas ≈1.2 ×1011M , baryon mass per

unit length of Mgas / l ≈ 3×108Mkpc

−1 , total mass of Mtot ≈ 9.7 ×1011M , and total mass per

unit length of Mtot / l ≈ 2 ×109Mkpc

−1 (all assuming unit filling factor, and with errors as

estimated above). For filament 2, these parameters are quite similar (length 400 kpc,

Mgas ≈ 3.6 ×1011M , Mgas / l ≈ 8 ×10

8Mkpc−1 , Mtot ≈ 3×10

12M , and

Mtot / l ≈ 7 ×109Mkpc

−1 ). We regard the results for filament 2 more tentative because of the

uncertainties discussed above.

The spatial morphology of filament 1 and possibly filament 2 may resemble simulations

of cold accretion flows (Dekel et al. 2009). In particular we emphasize the low velocity

dispersion in the lines, suggesting relatively quiescent gas. Of course the gas has recently been

energized by the QSO. The impact of local ionization, heating, and radiation pressure may have

not yet had time to disrupt inflowing gas.

There is evidence that filament 1 is an inflow. Filament 1 has a wavelength near 4682Å.

The QSO systemic redshift is zsys = 2.843± 0.001or λsys = 4671.9 ±1.2Å. The systemic redshift

is estimated using the MgII emission line, various Balmer lines, the onset of the Ly alpha forest,

and the center of mass of the galaxies in its vicinity, all of which give the same redshift

of 62

zsys = 2.843± 0.001 (Trainor & Steidel 2012). Filament 1 is redshifted with respect to systemic

by 650 ± 80 km/s, an 8-sigma detection. There is also an absorption feature at this wavelength,

but it is dominated by a metal line complex, AlII 1671Å, associated with an intervening Lyα

absorber at z=1.802. If a local absorber is present it is masked by the intervening system. The

filament is clearly physically associated with the QSO.

Since the emission appears on the red side of the systemic QSO velocity, three models for

the filament are possible. The filament could be a tidal stream between the QSO host galaxy and

an interacting object. However, the length of filament 1 is as much as an order of magnitude

longer than typical tidal streams. The velocity shear between the filament and the QSO is also

very large at the base of the filament (Rampazzo et al. 2005).

Alternatively, if the gas is on the far side of QSO it could be outflowing. Extended

emission line nebula are common in radio-load QSOs (Heckman et al. 1991a; Heckman et al.

1991b). They however exhibit large velocity dispersions (σv>500 km/s), have scale lengths of

~20 kpc, and do not display the narrow morphology of filament 1. While cooling could in

principle lead to a negative velocity dispersion gradient with radius, this behavior the very

narrow line widths observed in the distant regions of filament 1 are atypical for any QSO or

galactic superwind ouflow. In addition, outflows (or jets) tend to be two-sided. There is no blue-

shifted component corresponding to filament 1. Filament 2 is nearly opposite, but has a velocity

near systemic. HS 1549+1919 is radio quiet. Radio quiet QSOs are also observed to have

extended nebulae (Christensen et al. 2006), with lower velocity dispersions (σv~200 km/s) and

large velocity shears with respect to the QSO systemic velocity. Their Lya luminosity is typically

~1% of the QSO luminosity, similar to our object. These do not display narrow filamentary

morphology, but previous observations have not had the sensitivity to reach the low flux levels

observed by PCWI. They are understood either as outflows or as inflows.

The inflow scenario is the most plausible and least contrived. From Figure 21 and Figure

28 we note that there is a significant rise in the velocity dispersion as the we move along the

filament toward the QSO. This rise is similar to that predicted by simulations (P. Hopkins,

private communication). Again, it is difficult to understand physically how an outflow would be

increasingly collimated and with decreasing velocity dispersion at larger distances, where the

dispersion drops to as low as 50 km/s. An outflow with a typical cone angle would exhibit a

of 62

velocity width larger than observed (comparable to the outflow velocity). The jump in velocity

dispersion at R=200 kpc is accompanied by a small redward shift in the velocity centroid of the

emission, suggesting an accelerating inward flow. This point is near the clump that is visible in

the narrow-band image, region a in Figure 1. The infall velocity of ~600 km/s (depending on the

angle to the line of site) is consistent with the high dark matter mass derived above for the

system. From the derived mass per unit length, and assuming an infall velocity of 600 km/s, we

would estimate a mass inflow rate of 200Myr−1 . Detailed geometric, kinematic, and radiative

transfer modelling, beyond the scope of this paper, is required to confirm that the observations

do not permit alternative interpretations.

5.3. Kinematic Properties

The PCWI instrument allows us to study the kinematics of the gas. First we calculate the

flux-weighted mean velocity and velocity dispersion of all the extended emission. The

distribution in velocity space relative to the QSO systemic velocity is given in Figure 30. The

mean is +10 km/s, within errors equal to the systemic velocity. The 1D line-of-sight velocity

dispersion is 700 km/s. This is large, corresponding to a 3D velocity dispersion of 1200 km/s,

and a halo mass of 1.5 ×1014M , and a factor of ~10 larger than the dark matter mass derived

above. We show below that this large velocity dispersion may be an artifact of the superposition

of virialized and unvirialized components.

We turn to the next highest moment. The velocity field is complex, but some order can be

discerned. Referring back to Figure 4, filament 2 is blueshifted toward the west, filament 1 is

redshifted toward the south-east, and there is tendancy for the other gas to follow this trend going

from blue to red. We can calculate a projected angular momentum, assuming that the gas is in

the plane of the sky (Note that if the motion were purely radial the total angular momentum

would be zero). The result is L ≈1016M km/s kpc , in the S-SW direction (assuming fV = 1).

This is consistent with a halo mass of 3×1013M which for a canonical spin parameter λ=0.05

would have an total (baryonic) angular momentum of 1.2 ×1017 2 ×1016( )M km/s kpc .

We showed earlier how intensity and density are approximately correlated even ignoring

distance effects (see Figure 27). Thus it is interesting to consider how the kinematics vary with

intensity and therefore density. We might expect that higher density gas is more likely to have

of 62

collapsed and virialized, while lower density gas may have yet to collapse. We show in Figure 31

the spectra obtained in intensity bins for the full field, along with Gaussian fits to the individual

kinematic components. In fact what is apparent is that at low intensities the spectrum is best fit

by multiple, separate, narrow components. As the intensity increases components drop out, and a

broad component appears centered on the QSO systemic velocity. This trend is summarized in

Figure 32 and Figure 33, strongly suggesting that the progression in intensity and density is

probing virialization resulting from heirarchical structure formation. The broad central

component has a 1D velocity dispersion σ v ≈ 350 − 400 km/s , corresponding to a circular

velocity of vc ≈ 600 − 700 km/s , and a virial mass of Mh ≈ 2 ×1013M , consistent with the gas

mass calculation above.

6. Summary

We have used PCWI to detect, map, and characterize the extended Lyα emission in the

vicinity of QSO1549+19. We summarize some of the main conclusions here.

1. Extensive emission is detected near the systemic velocity of the QSO, at intensities

consistent with expected Lyman α fluorescence in gas of physical densities of 10-3-10-1 cm-3,

corresponding to overdensities at this redshift of δ 102 −104 .

2. Extensive simulation tests show that the adaptive smoothing algorithm provides a

robust method to detect extended, low surface brightness, filamentary line emission in low to

moderate signal-to-noise spectral data cubes.

3. While the emission morphology is complex, clear evidence of filaments are present. In

particular a narrow, kinematically cold filament (#1) is detected east of the QSO and redshifted

~10Å with respect to the systemic QSO redshift. There is tentative evidence for a second

filamentary or web-like extension west of the QSO connecting the QSO to a previously detected

Lyman alpha emitter, which we denote filament 2. These filamentary extensions are detected as

far as 50 arcsec, or 400 kpc from the QSO.

4. The faintest regions of emission also display more quiescent kinematics, and both

spectral image plots and intensity-binned spectra show a clear trend for faint, distant emission to

be kinematically quiet, while emission nearer the QSO and brighter in intensity (and likely

density) shows higher velocity dispersions.

of 62

5. Evidence is found for an inflowing filament of cold gas (filament #1). The filament is

narrow, kinematically cold at the extremity, and progressively disturbed kinematics as it

approaches the QSO. As such it is consistent with cold flow accretion models of inflowing gas.

Alternate models are less compelling, but a definitive conclusion will require further modelling

and observations of additional lines to constrain gas metallicity, radiative transfer, and geometry.

6. A broad central component is detected with a velocity dispersion σ v ≈ 350 − 400 km/s ,

corresponding to a circular velocity of vc ≈ 600 − 700 km/s , and a total virial mass of

1013.4±0.5M .

7. A simple CLOUDY ionization model is used to deduce physical properties of the gas,

notably density and gas mass. Assuming unit filling factor, filament gas masses are

(1− 3)×1011M while the total gas mass is 3×10

12M , consistent with the kinematically

derived virial mass of 1013.4±0.5M . We find that to first order density and intensity are

monotonically related, independently of distance from the QSO. The absence of extensive

regions of double-peaked Lyα emission nominally expected from optically thick Lyα

fluorescence is a direct result of the bright radiation field which renders most clouds optically

thin in the Lyman continuum with only modest HI column densities.

8. The angular momentum of the emitting gas is shown also to be consistent with the

deduced halo mass and a typical spin parameter.

of 62

Acknowledgements

We thank Tom Tombrello and Shri Kulkarni for their support of PCWI. We thank Marty Crabill,

Steve Kaye and the staff of the Palomar Observatory for their constant support. Nicole Lingner

participated in the observations. We are deeply grateful to Dean Joe Shepard, to the Caltech

Counselling Office, and to the family of Daphne Chang for their strength and support. We

acknowledge the detailed and helpful comments from the anonymous referee. This work was

supported by the National Science Foundation and the California Institute of Technology.

of 62

Tables

Region Area

[Arcsec2]

Band Flux

1

Flux

2

Flux

3

Flux

4

Flux

5

Luminosity

[erg/s]

S/N

a 100 4660-4690Å 8.2e-16 8.2e-18 2.7e-19 81460 2630 5.7e+43 14.5

b 100 4655-4685Å 8.9e-16 8.9e-18 2.9e-19 87540 2820 6.1e+43 16.0

c 150 4660-4680Å 4.9e-16 3.3e-18 1.6e-19 32370 1540 3.4e+43 6.9

d 150 4655-4690Å 7.5e-16 5.0e-18 1.4e-19 49070 1360 4.4e+43 10.4

e 100 4660-4680Å 3.7e-16 3.7e-18 1.8e-19 36520 1740 5.1e+43 6.9

tot 4450 4660-4690Å 2.0e-13 3.6e-18 1.2e-19 35425 1140 1.3e+46 43.5

QSO 4660-4690Å 1.0e+48

Table 1. Derived Fluxes and Detection Significance for Regions Defined in Figure 1.

Flux 1 = erg cm-2 s-1. Flux 2 = erg cm-2 s-1 arcsec-2. Flux 3 = erg cm-2 s-1 arcsec-2 Å-1. Flux 4 =

LU. Flux 5 = LU/Å. Errors in each unit can be derived by dividing flux by S/N.

of 62

Figures

Figure 1. Images of QSO HS1549+19. a. Narrow band image of QSO HS1549+19. The image is obtained on LRIS at Keck 1, using a narrow band filter 80Å centered on 4670Å. The image is a co-added sum totaling 3 hours of exposure in the narrow-band, and the V-band image in Figure 16 is used to continuum-subtract the image. The image has been smoothed with a Gaussian kernel of radius 1.5 arcsec. Full scale (0.02) corresponds to a flux of 1.6e-17 erg cm-2 s-1 arcsec-2. The 1 sigma noise is ~2.5e-19 erg cm-2 s-1 arcsec-2 in a 1 arcsec2 box. The continuum subtraction error is estimated to be ~2.5% full scale or 4e-19 erg cm-2 s-1 arcsec-2. We show a number of regions as labeled yellow boxes. The first four (a-d) are distinct “blob” features in the narrow-band image. Each has a LRIS spectroscopic observation and an associated mean redshift.

of 62

b. Corresponding PCWI image, centered on 4670Å, 40Å wide total slice using 3DASL images (see text). The color bar gives a linear intensity scale with 100,000 LU full range. 100,000 LU = 100,000 ph cm-2 s-1 sr-1 = 10-17 erg cm-2 s-1 arcsec-2. Smoothing is performed using 3DASL with a S/N threshold of 2.5 after masking the QSO within a 5 arcsec radius. Blobs a, b, and c are detected in PCWI image, albeit at lower spatial resolution due to the 2.4 arcsec slice sampling resolution. Blobs d are not clearly detected. Note there are some discrepant regions when comparing panels a. and b. One to one correspondence is not expected because of the lower angular resolution of PCWI, the higher noise in a wide PCWI slice (due to lower throughput, smaller telescope aperture, and higher sky background), and the continuum subtraction error in the narrow-band image. We show in Figure 2 that the two data sets are reasonably consistent.

of 62

c. Smoothed spectral images of QSO HS1549+19, 4Å wide slice, centered on 4672Å. The scale in arcsec is centered on the QSO. The color bar gives a linear intensity scale with 15000 LU full range. 10000 LU = 10000 ph cm-2 s-1 sr-1 = 10-18 erg cm-2 s-1 arcsec-2. Smoothing is performed using 3DASL with a S/N threshold of 2.5 after masking the QSO within a 5 arcsec radius. Note that the image looks quite different from that in panel b, a result of the much smaller wavelength range near the systemic velocity. In these images we can identify one possible ‘filamentary’ extension, connecting regions b, e, and c. We tentatively denote this feature “Filament 2”.

of 62

d. Smoothed spectral images of QSO HS1549+19, 4Å wide slice, centered on 4681Å. The scale in arcsec is centered on the QSO. The color bar gives a linear intensity scale with 15000 LU full range. 10000 LU = 10000 ph cm-2 s-1 sr-1 = 10-18 erg cm-2 s-1 arcsec-2. Smoothing is performed using 3DASL with a S/N threshold of 2.5 after masking the QSO within a 5 arcsec radius. Note that the image looks quite different from that in panel b. What appeared to be a distinct blob in panel a and b manifests as a linear extension eastward, which we denote “Filament 1”.

of 62

Figure 2. Comparison of flux in regions shown in Figure 1 from the Keck narrow-band image and PCWI shows good correspondence. Both fluxes are in units of erg cm-2 s-1 arcsec-2 averaged in the regions. Error bars for the narrow-band image include estimated continuum subtraction error of ~8 x 10-19 erg cm-2 s-1 arcsec-2. Dotted line shows 1 to 1, solid line shows linear fit. Flux offset is 4.7 x 10-19 erg cm-2 s-1 arcsec-2. Fit chi-square is 12.8 for 13 degrees of freedom.

of 62

Figure 3. Spectra from unsmoothed data cube in regions shown in Figure 1. Units are LU/Å. Error bars are calculated from the variance cube. Red lines show measured redshift based on LRIS spectra. For region e there is no LRIS spectrum, and we show Lyα at the systemic velocity. Regions a and e have quite narrow line widths, region c intermediate, and regions b and d broad. The final panel, ‘tot’, shows the total spectrum from the QSO region with the QSO masked. Detected fluxes and significance are presented in Table 1.

of 62

Figure 4. Smoothed spectral images of QSO HS1549+19 for several wavelengths. Each panel consists of a 4Å wide slice, centered on a) 4656Å, b) 4660Å, c) 4664Å, d) 4668Å, e) 4672Å, f) 4676Å, g) 4680Å, h) 4684Å, and i) 4688Å. The scale in arcsec is centered on the QSO. The color bar gives a linear intensity scale with 15000 LU full range. 10000 LU = 10000 ph cm-2 s-1 sr-1 = 10-18 erg cm-2 s-1 arcsec-2. Smoothing is performed using 3DASL with a S/N threshold of 2.5 after masking the QSO within a 5 arcsec radius. In these images we can identify at least two ‘filamentary’ extensions. One extends to the south-east in the 4680-4684Å slices, and we denote this “Filament 1”. The other connects the QSO with the clump at 4672Å, Δα=-40, Δδ=-7 arcsec that is evident in Figure 1, which we denote “Filament 2.

of 62

Figure 5. Comparison of raw and smoothed slices with 3DASL images. Top row: 4671Å, bottom row: 4682Å. Left: raw difference cube, 4Å slice. Middle: raw difference cube, 4Å slice, smoothed by 9 pixels. Right: 3DASL with S/N threshold 2.5. Note that image scales, given by color bars, are different for each image type.

Figure 6. Smoothed 1Å slices using 3DASL covering band around QSO systemic velocity,

4672Å. S/N threshold is 2.5. Filament 2 is detected in at least 4 slices.

of 62

Figure 7. Smoothed 1Å slices using 3DASL covering band around 4682Å. S/N threshold is 2.5. Filament 1 is detected in at least 7 slices.

Figure 8. Simulated 2D image field with sources only, illustrating several realizations of 2DASL compared to simple smoothing. a. simulated field, no noise. b. simulated field, with noise (one realization, σ=0.2). c. average of 10 2DASL realizations (example realizations shown in e-h). d. average of 10 simple 30 pixel smooth realizations. e, f, g, h. 4 examples of realizations of 2DASL. Comparison of panels c and d show that simple smoothing generates spurious filamentary features from compact sources distributed in a filamentary pattern, while 2DASL does avoids this problem.

of 62

Figure 9. Simulated 2D image field with sources and a filament, illustrating several realizations of 2DASL compared to simple smoothing, using a S/N threshold of 2.5. The filament is a 2D Gaussian with FWHM 20 pixels by 300 pixels, with a mean surface brightness (within FWHM) of 0.029. a. simulated field, no noise. b. simulated field, with noise (one realization, σ=0.2). c. average of 10 2DASL realizations (e-h?). d. average of 10 simple 30 pixel smooth realizations. e, f, g, h. 4 examples of realizations of 2DASL. As in Figure 8, comparison of panels c and d show that simple smoothing generates spurious filamentary features from compact sources distributed in a filamentary pattern, while 2DASL does avoids this problem. Comparison of panels c and d show that simple smoothing generates spurious filamentary features from compact sources distributed in a filamentary pattern, while 2DASL does avoids this problem.

of 62

Figure 10. Simulated 2D image field with sources and a filament, illustrating several realizations of 2DASL compared to simple smoothing, using a S/N threshold of 3.0. The filament is a 2D Gaussian with FWHM 20 pixels by 300 pixels, with a mean surface brightness (within FWHM) of 0.029. a. simulated field, no noise. b. simulated field, with noise (one realization, σ=0.2). c. average of 10 2DASL realizations (e-h). d. average of 10 simple 30 pixel smooth realizations. e, f, g, h. 4 examples of realizations of 2DASL. See comments in Figure 8 and Figure 9.

of 62

Figure 11. Simulated 2D image field with sources and a filament, illustrating several realizations of 2DASL compared to simple smoothing, using a S/N threshold of 2.0. The filament is a 2D Gaussian with FWHM 20 pixels by 300 pixels, with a mean surface brightness (within FWHM) of 0.029. a. simulated field, no noise. b. simulated field, with noise (one realization, σ=0.2). c. average of 10 2DASL realizations (e-h). d. average of 10 simple 30 pixel smooth realizations. e, f, g, h. 4 examples of realizations of 2DASL. See comments in Figure 8 and Figure 9.

of 62

Figure 12. Simulated 2D image field with sources (including a very bright source at the center, approximately 10,000 times the surface brightness of the filament, and not masked), and a filament with a narrow and broad component, illustrating several realizations of 2DASL compared to simple smoothing, using a S/N threshold of 2.5. a. simulated field, no noise. b. simulated field, with noise (one realization, σ=0.2). c. average of 10 2DASL realizations (e-h). d. average of 10 simple 30 pixel smooth realizations. e, f, g, h. 4 examples of realizations of 2DASL. See comments in Figure 8 and Figure 9.

of 62

Figure 13. Simulated 2D image field, showing a series of filaments with increasing intensity. Each filament is a 2D Gaussian with FWHM 20 pixels by 300 pixels. Mean surface brightness (within FWHM) from left to right is 0.014, 0.021, 0.029 (same as Figures 7-9) Figure 9, 0.036, 0.043, 0.057. a. simulated field, no noise. b. simulated field, with noise (one realization, σ=0.2). c. average of 10 2DASL realizations (d-g). d. average of 10 simple 30 pixel smooth realizations. e, f, g, h. 4 examples of realizations of 2DASL. 2DASL realizations use S/N threshold of 2.5. Filament 3 (average brightness 0.029) is considered barely detected, as expected since the signal-to-noise ratio is approximately 0.029/(0.2/FWHM)=0.029/(0.2/20)=2.9.

of 62

Figure 14. Smoothed spectral images (1Å slices) for showing the effect of varying the S/N threshold of the algorithm. a. S/N 3.0, 4671Å. b. S/N 3.0, 4681Å. c. S/N 2.5, 4671Å. d. S/N 2.5, 4681Å. e. S/N 2.0, 4671Å. f. S/N 2.0, 4681Å.

of 62

Figure 15. Raw, smoothed, and 3DASL images of data (top row) in 1Å slice, and pure noise (bottom row) obtained using actual pipeline and image positions in final mosaic, also a 1Å slice. Left: raw image mosaic. Middle: raw mosaic smoothed by 9 x 9 boxcar. Right: 3DASL image, S/N threshold 3.0, 1Å wide. Note that because of the 3D smoothing, some features that appear in a single 1Å slice will not be detected if they are not correlated from slice to slice. For example, in the data (top row), emission appears at Δα=-30, Δδ=-8, but is absent in the 3DASL image on the right.

Figure 16. Raw, smoothed, and 3DASL images of simulated data cube with single line filament similar to observed filament 1, FWHM 3Å, fixed center (4682Å). Upper row: 4Å slices. Lower row: 1Å slices. Left: raw image mosaic. Middle: raw mosaic smoothed by 9 x 9 boxcar. Right: 3DASL image, S/N threshold 3.0.

of 62

Figure 17. Simulated data cube with single line filament, smoothed with 3DASL, cut in spectral-image direction along filament, S/N threshold 3.0. a. No velocity change along filament. b. velocity change ~600 km/s over filament.

Figure 18. Simulated filament in western direction of comparable intensity to filament 2. This simulation demonstrates that the detectability of filamentary emission is not affected by the lack of north-south dithering present in the western portion of the data cube mosaic. a. Image slice centered on 4672Å, 4Å wide from 3DASL cube at S/N>3.0. b. Spectral image plot at azimuth -110°. Filament has total intensity 18,750LU, width 12 arcsec, velocity dispersion 60 km/s.

of 62

Figure 19. Illustration of location and impact of compact sources. a. V-band continuum image showing all compact sources with V<25.0. Yellow boxes show continuum-only sources, and red boxes show Lya emitters. Boxes denote region replaced with local sky on a wavelength by wavelength basis. Sky is obtained using an annular zone. b. Narrow-band image with source boxes. c. 3DASL smoothed 1Å slice at 4682Å after source removal. d. 3DASL smoothed 1Å slice at 4682Å with no source removal. There is no significant difference between the source-removed and non-sourced-removed images in this and at all wavelengths.

of 62

Figure 20. A set of spectral-image plots at different azimuthal positions centered on the QSO. Each plot is obtained in a pseudo-slit 5 arcsec wide. R000 is 0 degrees from the horizontal axis towards East. Angle is clockwise from this, so that for example 90 degrees corresponds to a North-South pseudo-slit. Vertical lines corresponds to Lyman α at QSO systemic redshift.

of 62

Figure 21. Smoothed image slices with delineated filament regions for filament 1 and spectral-image plot. a: 3DASL smoothed image slice, 4Å wide, S/N threshold 2.0 to bring out faintest extension of. b. Spectral image plot, summed over pseudo slit shown in panel a., full scale 5000 LU/Å, ordinate is distance along region in arcsec from QSO. There is a suggestion of a double-peaked line profile in the range Δθ~20-40 arcsec.

Figure 22. Smoothed image slices with delineated filament regions for filament 2 and spectral-image plot. The path of the pseudo-slit chosen to delineate filament 2 is not unique, and other choices produce similar results. a: 3DASL smoothed image slice, 4Å wide, S/N threshold 2.5. b. Spectral image plot, summed over pseudo slit shown in panel a., full scale 5000 LU/Å, ordinate is distance along region in arcsec from QSO.

of 62

Figure 23. Spectrum of QSO HS1549+19 in PCWI spectral range. Absorption lines are present at 4640.5, 4643.9, 4648.3, 4654.4, 4661.9, 4668.9, and 4981.3Å.

of 62

Figure 24. Inferred cloud density vs. radius and intensity from CLOUDY model. Dots give measured intensity vs. radius. Colored lines give predicted intensity vs. radius for given densities shown in legend. Solid line gives total intensity, and dotted line gives line-pumped intensity. Only for the highest densities at large radii are significant Lyman α fluorescence and corresponding double-peaked emission from an optically thick source expected. All model results are for volume filling factor unity ( fV = 1) and cloud thickness 40 kpc. Yellow vertical line indicates radii within which QSO masking could impact results.

of 62

Figure 25. Inferred cloud line center optical depth vs. cloud density. Points display intensities converted into line center optical depth based on model HI column density and temperature. All model results are for volume filling factor unity ( ) and cloud thickness 40 kpc. Yellow vertical line indicates radii within which QSO masking could impact results.

fV = 1

of 62

Figure 26. Temperature vs. radial distance for complete set of cloud densities, CLOUDY model. Colors indicate density. All model results are for volume filling factor unity ( ) and cloud thickness 40 kpc. Cross symbols indicate mean temperatures inferred for the data vs. radius. The temperatures are high in the CQM, compared to typical photoionized gas, because the QSO is hyperluminous, and the ionization parameter U>10.

fV = 1

of 62

Figure 27. Inferred density given surface brightness from CLOUDY model. Line show for the intrinsic cloud density vs. the observed intensity for each radial position from the QSO. Radii range from 30 kpc to 600 kpc (red to violet) in equal logarithmic steps. Without knowing the physical radial distance, the error in the conversion from intensity to density is ~0.3 dex. Right axis shows the overdensity parameter for standard baryon fraction at this redshift. All curves assume unit volume filling factor ( fV = 1) and cloud thickness 40 kpc.

of 62

Figure 28. Inferred physical properties along filament 1. All curves assume unit volume filling factor ( fV = 1 ) and cloud thickness 40 kpc. The vertical dotted line at Δθ = 220 kpc corresponds to the virial radius based on the mass estimate. The vertical dashed line indicates the inner region near QSO that has been masked. The horizontal dotted line in the velocity (bottom) plot indicates the QSO systemic velocity.

of 62

Figure 29. Inferred physical properties along filament 2. All curves assume unit volume filling factor ( fV = 1 ) and cloud thickness 40 kpc. The vertical dashed line indicates the inner region near QSO that has been masked. The horizontal dotted line in the velocity (bottom) plot indicates the QSO systemic velocity.

of 62

Figure 30. Velocity profile of intensity, along with Gaussian fit showing 1D line-of-sight velocity dispersion σσv=700 km/s.

of 62

Figure 31. Spectra of emission binned by intensity. This figure shows the normalized spectrum binned by intensity (LU/Å) as shown in each panel, using the smoothed data cube. The spectra are fit with multiple Gaussian distributions, shown in red. Line centroid and velocity width from fits are used in the following figure. The fraction of the estimated total mass (see text) corresponding to each intensity bin is shown after the intensity range.

of 62

Figure 32. Summary of kinematic components vs. intensity (density). The spectral line fits to the intensity binned spectra shown in Figure 31 are displayed in this plot vs. intensity (lower scale), and overdensity (upper scale, see Figure 27). Colors are used to separate distinct kinematic components. Circle location in velocity is best fit central velocity for each component. Vertical bar shows +/- 1-sigma velocity width of Gaussian fit. Size of dot indicates fraction of total emission in this component (from Gaussian height fit). Horizontal lines at left show QSO absorption line components.

of 62

Figure 33. Summary of kinematic components vs. intensity (density) with overlay. The spectral line fits to the intensity binned spectra shown in Figure 31 are displayed in this plot vs. intensity (lower scale), and overdensity (upper scale). Colors are used to separate distinct kinematic components. Circle location in velocity is best fit central velocity for each component. Vertical bar shows +/- 1-sigma velocity width of Gaussian fit. Size of dot indicates fraction of total emission in this component (from Gaussian height fit). Horizontal lines at left show QSO absorption line components. In this version of the figure we have overlaid a cartoon (shaded regions) suggesting that higher densities are kinematically evolved in the sense of harboring fewer and broader kinematic components. The implication is that the material at higher density is more virialized.

of 62

References

Adelberger, K. L., Steidel, C. C., Kollmeier, J. A., & Reddy, N. A. 2006, The Astrophysical Journal, 637, 74 Binette, L., Wang, J. C. L., Zuo, L., & Magris, C. G. 1993, The Astronomical Journal, 105, 797

Birnboim, Y., & Dekel, A. 2003, Monthly Notices of the Royal Astronomical Society, 345, 349 Bregman, J. N. 2007, Annual Review of Astronomy and Astrophysics, 45, 221

Cantalupo, S., Lilly, S. J., & Haehnelt, M. G. 2012, Monthly Notices of the Royal Astronomical Society, 425, 1992

Cantalupo, S., Porciani, C., Lilly, S. J., & Miniati, F. 2005, The Astrophysical Journal, 628, 61 Cen, R., Miralda-Escudé, J., Ostriker, J. P., & Rauch, M. 1994, The Astrophysical Journal, 437, L9 Cen, R., & Ostriker, J. P. 2006, The Astrophysical Journal, 650, 560

Christensen, L., Jahnke, K., Wisotzki, L., & Sanchez, S. F. 2006, Astronomy and Astrophysics, 459, 717

Dekel, A., et al. 2009, Nature, 457, 451 Dijkstra, M., Stuart, J., & Wyithe, B. 2006, Monthly Notices of the Royal Astronomical Society, 372, 1575 Fardal, M. A., Katz, N., Gardner, J. P., Hernquist, L., Weinberg, D. H., & Davé, R. 2001, The Astrophysical Journal, 562, 605 Faucher-Giguère, C.-A., & Kereš, D. 2011, Monthly Notices of the Royal Astronomical Society, 412, L118 Faucher-Giguère, C.-A., Kereš, D., Dijkstra, M., Hernquist, L., & Zaldarriaga, M. 2010, The Astrophysical Journal, 725, 633 Faucher-Giguère, C.-A., Kereš, D., & Ma, C.-P. 2011, Monthly Notices of the Royal Astronomical Society, 417, 2982 Ferland, G. J. 1996, University of Kentucky Department of Physics and Astronomy Internal Report Förster Schreiber, N. M., et al. 2006, The Astrophysical Journal, 645, 1062

Francis, P. J., & Bland-Hawthorn, J. 2004, Monthly Notices of the Royal Astronomical Society, 353, 301

Fukugita, M., & Peebles, P. J. E. 2004, The Astrophysical Journal, 616, 643 Fumagalli, M., Prochaska, J. X., Kasen, D., Dekel, A., Ceverino, D., & Primack, J. R. 2011, Monthly Notices of the Royal Astronomical Society, 418, 1796 Furlanetto, S. R., Schaye, J., Springel, V., & Hernquist, L. 2003, Astrophysical Journal, 599, L1

---. 2005, Astrophysical Journal, 622, 7

of 62

Fynbo, J. U., Moller, P., & Warren, S. J. 1999, Monthly Notices of the Royal Astronomical Society, 305, 849

Gould, A., & Weinberg, D. H. 1996, Astrophysical Journal v468, 468, 462 Gunn, J. E., & Peterson, B. A. 1965, The Astrophysical Journal, 142, 1633

Haiman, Z., & Rees, M. J. 2001, The Astrophysical Journal, 556, 87 Haiman, Z., Spaans, M., & Quataert, E. 2000, The Astrophysical Journal, 537, L5

Hayashino, T., et al. 2004, The Astronomical Journal, 128, 2073 Heckman, T. M., Lehnert, M. D., Miley, G. K., & van Breugel, W. 1991a, The Astrophysical Journal, 381, 373 Heckman, T. M., Miley, G. K., Lehnert, M. D., & van Breugel, W. 1991b, The Astrophysical Journal, 370, 78 Hennawi, J. F., & Prochaska, J. X. 2013, The Astrophysical Journal, 766, 58

Hennawi, J. F., Prochaska, J. X., Kollmeier, J., & Zheng, Z. 2009, The Astrophysical Journal, 693, L49

Hernquist, L., Katz, N., Weinberg, D. H., & Miralda-Escudé, J. 1996, Astrophysical Journal Letters v457, 457, L51

Hogan, C. J., & Weymann, R. J. 1987, Monthly Notices of the Royal Astronomical Society, 225, 1P

Hu, E. M., & Cowie, L. L. 1987, The Astrophysical Journal, 317, L7 Kereš, D., Katz, N., Weinberg, D. H., & Davé, R. 2005, Monthly Notices of the Royal Astronomical Society, 363, 2 Kirkman, D., & Tytler, D. 2008, Monthly Notices of the Royal Astronomical Society, 391, 1457

Kollmeier, J. A., Zheng, Z., Davé, R., Gould, A., Katz, N., Miralda-Escudé, J., & Weinberg, D. H. 2010, The Astrophysical Journal, 708, 1048

Komatsu, E., et al. 2011, The Astrophysical Journal Supplement Series, 192, 18 Lynds, R. 1971, The Astrophysical Journal, 164, L73

Matuszewski, M., Rahman, S., Morrissey, P., Moore, A., & Martin, D. C. 2010, Proc SPIE, Ground-based and Airborne Instrumentation for Astronomy III, 7735

Meiksin, A. A. 2009, Reviews of Modern Physics, 81, 1405 Nelson, D., Vogelsberger, M., Genel, S., Sijacki, D., Keres, D., Springel, V., & Hernquist, L. 2013, Monthly Notices of the Royal Astronomical Society, 429, 3353 Nilsson, K. K., Fynbo, J. P. U., Moller, P., Sommer-Larsen, J., & Ledoux, C. 2006, Astronomy and Astrophysics, 452, L23 North, P. L., Courbin, F., Eigenbrod, A., & Chelouche, D. 2012, Astronomy and Astrophysics, 542, A91

of 62

Prochaska, J. X., & Tumlinson, J. 2009, in adsabsharvardedu (Dordrecht: Springer Netherlands), 419

Rampazzo, R., Plana, H., Amram, P., Bagarotto, S., Boulesteix, J., & Rosado, M. 2005, Monthly Notices of the Royal Astronomical Society, 356, 1177

Rauch, M., Becker, G. D., Haehnelt, M. G., Gauthier, J. R., & Sargent, W. L. W. 2013, Monthly Notices of the Royal Astronomical Society, 429, 429

Rauch, M., et al. 2008, The Astrophysical Journal, 681, 856 Rauch, M., et al. 1997, Astrophysical Journal v489, 489, 7

Sembach, K. R., & Tonry, J. L. 1996, Astronomical Journal v112, 112, 797 Steidel, C. C., Adelberger, K. L., Shapley, A. E., Pettini, M., Dickinson, M., & Giavalisco, M. 2000, Astrophysical Journal, 532, 170 Steidel, C. C., Bogosavljević, M., Shapley, A. E., Kollmeier, J. A., Reddy, N. A., Erb, D. K., & Pettini, M. 2011, The Astrophysical Journal, 736, 160 Trainor, R. F., & Steidel, C. C. 2012, The Astrophysical Journal, 752, 39

van de Voort, F., Schaye, J., Booth, C. M., & Dalla Vecchia, C. 2011a, Monthly Notices of the Royal Astronomical Society, 415, 2782

van de Voort, F., Schaye, J., Booth, C. M., Haas, M. R., & Dalla Vecchia, C. 2011b, Monthly Notices of the Royal Astronomical Society, 414, 2458

Zhang, Y., Anninos, P., & Norman, M. L. 1995, The Astrophysical Journal, 453

igm emission observations with the cosmic web …authors.library.caltech.edu/44109/1/1402.4816.pdf2....

Documents